Active learning based target detection model training method

Abstract

The present invention provides a method for training a target detection model based on active learning, comprising: inputting a sample image marked with an expert label from an expert-calibrated input data set to a model training data set, and / or inputting a model training method from an uncalibrated input data set The data set is input without a sample image of the target object; when the number of sample images in the model training data set is less than or equal to the upper limit of the data, continue to train the initial model; when the number of sample images is greater than the upper limit of the data, select the filter After removing the image, train the initial model, select the input data set of difficult samples without expert labels in the filtered image; perform target object frame selection screening and expert label marking on the filtered image, and then input the expert calibration input data set, Repeat the above scheme to generate a target detection model. The invention solves the problem in the prior art that in order to improve the generalization ability of the target detection model, the accumulation of the data volume of the training data set results in low model training efficiency.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for training a target detection model based on active learning. Background technique [0002] Target detection is an image understanding algorithm based on target geometric and statistical features. Target detection combines the positioning and recognition of target objects. For example, based on computer vision algorithms, different types of target objects in the image are detected, that is, by The rectangular frame marks the location of the target and identifies the category of the target object. [0003] In order to make the target detection model suitable for different environmental scenarios and improve the generalization ability of the target detection model, in the process of training the target detection model, the model parameters are usually continuously optimized and adjusted to achieve the purpose of model refinement; this requires Periodically inpu...

AI Technical Summary

Problems solved by technology

[0004] The main purpose of the present invention is to provide a target detection model training method based on active learning to solve the problem of the accumulation of data in the training data set in order to improve the generalization ability of the model during the training process of the target detection model in the prior art. The problem of low training efficiency

Images